Jev AI for trading: useful probabilities, not a profit machine.

Jev can evaluate a shared market state against several bounded questions in parallel. That is useful for research workflows where every answer, threshold, and later outcome must be visible.

Decisions shown on the public page are labelled demonstrations. Live evaluation is verified on the server and kept deliberately separate from the public demo.

What this project tests

Whether a particular combination of live state, typed probabilities, confidence threshold, and simulated holding period produces evidence worth further study after costs. It does not claim that a model predicts markets or recommend a trade.

Why paper trading comes first

Live price displays are not evidence. The ledger records a decision timestamp and price, then samples the later market state so the project can calculate hit rate and directional movement rather than selecting attractive screenshots.

What is intentionally excluded

No autonomous real-money execution, custody, personalised financial advice, or claims of expected returns.